netdev - Re: [PATCH] Netlink messages for multicast HW addr programming

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20160203142250.GB2219@nanopsycho.orion>
Date:	Wed, 3 Feb 2016 15:22:50 +0100
From:	Jiri Pirko <jiri@...nulli.us>
To:	Patrick Ruddy <pruddy@...cade.com>
Cc:	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	David Stapleton <dstaplet@...cade.com>,
	Olufemi Komolafe <okomolaf@...cade.com>,
	Stephen Hemminger <shemming@...cade.com>
Subject: Re: [PATCH] Netlink messages for multicast HW addr programming

Wed, Feb 03, 2016 at 03:14:41PM CET, pruddy@...cade.com wrote:
>On Sat, 2016-01-23 at 08:53 +0100, Jiri Pirko wrote:
>> Fri, Jan 22, 2016 at 04:49:36PM CET, pruddy@...cade.com wrote:
>> >On Fri, 2016-01-22 at 16:05 +0100, Jiri Pirko wrote:
>> >> Fri, Jan 22, 2016 at 03:51:55PM CET, pruddy@...cade.com wrote:
>> >> >On Thu, 2016-01-21 at 21:04 +0100, Jiri Pirko wrote:
>> >> >> Thu, Jan 21, 2016 at 12:47:54PM CET, pruddy@...cade.com wrote:
>> >> >> >Add RTM_NEWADDR and RTM_DELADDR netlink messages to indicate
>> >> >> >interest in specific multicast hardware addresses. These messages
>> >> >> >are sent when addressed are added or deleted from the appropriate
>> >> >> >interface driver.
>> >> >> >
>> >> >> >Signed-off-by: Patrick Ruddy <pruddy@...cade.com>
>> >> >> >
>> >> >> >diff --git a/net/core/dev_addr_lists.c b/net/core/dev_addr_lists.c
>> >> >> >index c0548d2..a0ebadd 100644
>> >> >> >--- a/net/core/dev_addr_lists.c
>> >> >> >+++ b/net/core/dev_addr_lists.c
>> >> >> >@@ -12,6 +12,7 @@
>> >> >> >  */
>> >> >> > 
>> >> >> > #include <linux/netdevice.h>
>> >> >> >+#include <net/netlink.h>
>> >> >> > #include <linux/rtnetlink.h>
>> >> >> > #include <linux/export.h>
>> >> >> > #include <linux/list.h>
>> >> >> >@@ -661,6 +662,62 @@ out:
>> >> >> > }
>> >> >> > EXPORT_SYMBOL(dev_mc_add_excl);
>> >> >> > 
>> >> >> >+static int fill_addr(struct sk_buff *skb, struct net_device *dev,
>> >> >> >+		     const unsigned char *addr, int type)
>> >> >> >+{
>> >> >> >+	struct nlmsghdr *nlh;
>> >> >> >+	struct ifaddrmsg *ifm;
>> >> >> >+
>> >> >> >+	nlh = nlmsg_put(skb, 0, 0, type, sizeof(*ifm), 0);
>> >> >> >+	if (nlh == NULL)
>> >> >> >+		return -EMSGSIZE;
>> >> >> >+
>> >> >> >+	ifm = nlmsg_data(nlh);
>> >> >> >+	ifm->ifa_family = AF_UNSPEC;
>> >> >> >+	ifm->ifa_prefixlen = 0;
>> >> >> >+	ifm->ifa_flags = IFA_F_PERMANENT;
>> >> >> >+	ifm->ifa_scope = RT_SCOPE_LINK;
>> >> >> >+	ifm->ifa_index = dev->ifindex;
>> >> >> >+	if (nla_put(skb, IFA_ADDRESS, dev->addr_len, addr))
>> >> >> >+		goto nla_put_failure;
>> >> >> >+	nlmsg_end(skb, nlh);
>> >> >> 
>> >> >> How can userspace tell if this is a multicast addr? Also, why not add
>> >> >> notifications for unicast addrs so we handle both the same?
>> >> >
>> >> >I am only using this RTM_NEW/DELADDR AF_UNSPEC message for multicast
>> >> >addresses. I understand that the unicast addresses are notified in the
>> >> >RTM_NEW/DELLINK.
>> >> 
>> >> dev->dev_addr is notified by that. UC addresses do not have any
>> >> notification now.
>> >It is the multicast MAC address (ie the multicast equivalent of the
>> >dev_addr) that we are notifying here right? Not L3 addresses.
>> 
>> I know :)
>> 
>> For example if kernel code calls dev_uc_add or dev_uc_del, thise
>> addresses are not notified to the userspace. All I'm saying is, if you
>> add notification for dev_mc_add and dev_mc_del, add it for uc variants
>> as well. Then you have to distinguish mc and uc in the netlink message.
>
>One more Q on this. I can use IFA_ADDRESS for unicast addresses and
>IFA_MULTICAST for multicast addresses but are these distinctions
>actually required since the multicast nature of an address is clear from
>the address itself - at least it is for ethernet and ifininet?

Please mark them correctly. Thanks.